Dear Sourena and Quiting,

Thank you Sourena, that was “sort of” my point.

More specifically: topup calculates the field by looking at the difference in location of the signal from two different scans with different blip direction. In some areas that difference is small, and there the (off-resonance) field is small. In other areas the difference is bigger and there the field is bigger.

In a GE-EPI the signal is weakened (or almost gone) in areas of strong (off-resonance field). That means that in the areas where we particularly want to be able to calculate the field we cannot, because if the signal is gone we cannot assess any difference in location of that (non existent) signal. Also, the level of signal loss can be different between the two blip-directions.

Hence, even if you want to ultimately apply your correction to GE-EPI (FMRI) you might want to precede that with a blipped pair of SE-EPI for calculating the field from. It is then best if all other acquisition parameters are identical between the SE-EPI and the GE-EPI.

Jesper

On 14 Feb 2014, at 03:59, Sourena Soheili <[log in to unmask]> wrote:

> I think Jesper was making this point: In se EPI the 180deg refocusing
> pulse (which is absent in ge EPI) effectively compensates for
> dephasing of spins in high susceptibility regions. This type of signal
> loss which happens in ge EPI is impossible to recover in postproc (but
> can be diminished by better shimming, slice tilting, shorter TE, etc).
> What topup effectively corrects is the spatial distortion stemming
> from the non-linearity of phase encoding blips which is imposed by
> magnetic susceptibility (in contrast to Eddy-current induced
> off-resonance which is inconsistent volumewise).
> 
> ge EPI should have more SNR than se EPI.
> 
> Sourena
> 
> On 2/14/14, Qiuting Wen <[log in to unmask]> wrote:
>> Dear Jesper,
>> 
>> In my experiment I did see that TOPUP does a better job in SE-EPI than
>> GE-EPI. Could you explain the signal dropout a little bit more? Is it
>> because GE-EPI is T2* weighted so it has worse SNR? And could it be fixed
>> if we acquire a couple more blipped passes then average to achieve a higher
>> SNR?
>> 
>> Thanks ahead.
>> 
>> Qiuting
>> 
>> 
>> On Tue, Feb 11, 2014 at 6:33 AM, Jesper Andersson <
>> [log in to unmask]> wrote:
>> 
>>> Dear Sourena,
>>> 
>>>> Thank you for explaining this issue.
>>>> As I have seen in HCP protocol the common practice is a pair of AP/PA
>>>> spin echo EPI with matched BW/FOV/matrix. Is it advisable to acquire
>>>> extra volumes?
>>> 
>>> Acquiring 2-3 volumes per blip-direction is very quick and offers some
>>> protection against something going one with one volume.
>>> 
>>>> or even some BOLD volumes with an opposing
>>>> phase-encoding direction?
>>> 
>>> BOLD (ge-epi) is worse for this since we then have signal dropout in
>>> addition to the distortions. If you acquire your ge- and se-EPIs with the
>>> same BW/FOV/matrix you can still use the field from the se-EPI for the
>>> ge-EPI data.
>>> 
>>> Jesper
>>> 
>>> 
>>>> 
>>>> Yes I was thinking about the gradual heating of the gradients. I
>>>> wonder if disturbance in the active shim is something serious to
>>>> avoid.
>>>> Sincerely yours,
>>>> Sourena
>>>> 
>>>> On 2/11/14, Jesper Andersson <[log in to unmask]> wrote:
>>>>> Dear Sourena,
>>>>> 
>>>>>> Should the topup approach be considered superior to fieldmap
>>>>>> correction regarding (single-band) fMRI?
>>>>> 
>>>>> if you are only getting a single pair for calculating a fieldmap I
>>>>> don't
>>>>> think it is superior to a "traditional" fieldmap. It is mainly just
>>> another
>>>>> way of obtaining the field.
>>>>> 
>>>>>> How serious is the effect of gradient thermal state on fMRI/DTI
>>>>>> signals specially when the scanner has just started its daily job?
>>>>>> Does there exist some guideline to adhere?
>>>>> 
>>>>> I am not sure what you mean by "gradient thermal state". Are you
>>>>> talking
>>>>> about the gradual heating of the gradients during a long run? If so, I
>>> don't
>>>>> think it is a problem for fMRI. Anything slow and gradual tend to be
>>>>> regressed out anyway.
>>>>> 
>>>>> For DTI I don't know. Maybe someone else has some experience?
>>>>> 
>>>>> Jesper
>>>>> 
>>>>>> 
>>>>>> Thanks in advance,
>>>>>> Sourena
>>>>> 
>>> 
>>